Embodied AI is an R&D project exploring how artificial intelligence can operate on human movement as a real-time, expressive input within live performance and immersive environments. The project investigates the integration of motion capture, 3D animation, and generative AI to create responsive systems where digital visuals and avatars are directly driven by human movement.

Developed by Alexander Whitley Dance Company, the project combines a custom pipeline using Unity, TouchDesigner, and Stable Diffusion to enable real-time interaction between performers and AI-generated imagery. This system was embedded into a live performance at Sadler’s Wells East, allowing the team to test both technical capabilities and audience perception in a real-world context.

The research highlights both the creative potential and current limitations of AI in this space—particularly the gap between real-time and offline generative systems, and the challenge of achieving expressive, responsive motion-driven outputs. The project positions AI not just as a tool, but as a new creative medium, opening pathways for future work in performance, immersive media, and interactive experiences.